The Droid Knight: a silent guardian for the Android kernel, hunting for rogue smartphone malware applications

Transcription

1 Virus Bulletin 2013 The Droid Knight: a silent guardian for the Android kernel, hunting for rogue smartphone malware applications Next Generation Intelligent Networks Research Center (nexgin RC) Institute of Space Technology (IST) National University of Computer and Emerging Sciences (NUCES) Islamabad, Pakistan 1

2 Authors - Introduction M Ali Akbar ( aliakbar@nexginrcorg) Research Engineer, nexgin RC, IST, Pakistan Farrukh Shahzad ( farrukhshahzad0@yahoocom) PhD Candidate, FAST-NU, Principal Malware Researcher, Ebryx (Pvt) Ltd Pakistan Muddassar Farooq ( muddassarfarooq@nexginrcorg) Director nexgin RC, Dean IST, Pakistan 2 2

3 Key points Motivation Malicious Applications? Challenges on Smartphone Types of Malware Detection In-execution Malware detection using PCB On Android Framework Empirical Evaluation Performance Overheads Conclusion 3 3

4 Motivation Market Share (1st Quarter 2012) - Gartner report 2012 Android Market Share : 641% iphone Market Share : 19% Symbian Market Share : 6% RIM Market Share : 59% Microsoft Market Share : 26% Market Share (2 nd Quarter 2013) - Gartner Report 2013 Android Market Share : 744% iphone Market Share : 182% Symbian Market Share : 06% Black Berry Market Share : 3% Microsoft Market Share : 29% Threats in Android - Malware samples Increased from 400 to (an increase of 3325%) 2011 [Juniper Report 2012] A sixfold increase in malicious applications ,000 [Trend-Micro 3 rd Quarter 2012] Malware Categories in Android [Juniper Report 2012] Spyware % SMS Trojan % SMS-Flooder % Worms % 4 4

5 Motivation (Android threat Growth) TREND MICRO: Mobile Threats and Security Round up Report

6 Motivation (Distribution of threats) TREND MICRO: Mobile Threats and Security Round up Report

7 Malicious Applications Worm A malicious code that usually spreads by exploiting vulnerabilities in the network services is termed as a worm eg ios Ikee Worm, Commwarrior etc Virus (only seen on windows mobiles) Viruses infect normal processes on a system and use them to execute their malicious code They usually spread through sharing of infected programs Duts etc Trojans They pose as legitimate and applications, execute their malicious codes in the background without user s knowledge eg ZeuS, SMSAndroidOSFoncy and SkullD etc Spyware Spyware present themselves as useful programs but their primary objective is to steal sensitive information about a user such as passwords, s etc eg GPSSMSSpy and Nickyspy RootKits Rootkits hide a malicious application on a system by modifying a system s kernel by modifying system calls are instrumented eg ITFUNZ and Z4Mod 7 7

8 High Detection Rate Low False Alarm Malware Detection on Smartphones Major Challenges Detection Delay (minimal) Phone s Performance Degradation (minimal) Robustness to Evasion Battery Constraints Processing Power Processor Architecture Issues (desktop solutions are not directly portable) 8 8

9 Types of Malware Detection Signature Based Malware Detection Non- Signature Based Signature Based: Detection on the basis of known byte sequences Unable to detect new malware Regular updates required Static Dynamic Non-Signature Based: Detection on the basis of smarter features After- Execution In-Execution Able to detect new malware Regular updates may not be necessary 9 9

10 Types of Malware Detection Signature Based Malware Detection Non- Signature Based Static Detection: Detection on the basis of file as residing on secondary storage Prone to techniques such as obfuscation and polymorphism Static After- Execution Dynamic In-Execution Dynamic Detection: Its detects malware on the basis of run-time behavior of a process (It s a more direct look) Resilient to techniques such as code obfuscation High processing overhead 10 10

11 Types of Malware Detection Malware Detection After-Execution Detection: Forensic and taint Analysis Offline analysis Signature Based Non- Signature Based Lower processing overhead Undo the Change made by it Static Dynamic In-Execution Detection: After- Execution In-Execution Detection during execution End user tool High processing overhead 11 11

12 Dynamic Analysis Techniques (on Android) Information Flow Tracking Dynamic Function Call Tracing Runtime Permissions Leak Detection Misbehavior analysis using power utilization patterns System Performance/Behavior based Anomaly Detection 13 13

13 Our Proposed Thesis To use parameters of Process Control Blocks in kernel of OS for malware detection Task_struct is Process control block in Android kernel Every executing process has an associated Task Structure in the kernel Task_struct is structure of multiple structures (along with it own native parameters) eg Memory Manager, Signal Structure, File system etc The parameter values of the TS fields are different for benign and malicious processes on the basis of their execution behavior So a collection of parameters can be use to discriminate between benign and malware processes Multi-Threaded applications launched from user space have a common task structure for all threads 16 16

14 Problem Statement To Identify a malicious process based upon its execution patterns in its process control blocks (using machine learning algorithms) and stop it before it can complete its execution 17 17

15 Requirements for our system High detection accuracy Low false alarm rate Low processing overhead Real-time classification Detect malware during execution Robust to evasion attempts 18 18

16 Datasets (Enhanced Results) Enhanced version of DroidKnight is presented as TStructDroid [Tech Rep] Malware Categories (110 Malware Samples) Trojans, SpyWare, BackDoor, Bots, RootKits Benign Categories (110 benign sample) Games, Image Viewers, Sketching tools, Text Editors, Image Editors, Android Utilities 19 19

17 Process Flow Diagram 20 20

18 Architecture of Framework Feature Extractor Java Apps Execution Kernel Module for PCB dump Features Analyzer Time-series Features Short-listing 21 21

19 contd Redundant feature-instance elimination Time-series Segmentation & Frequency Information Extraction Variance Calculation over time-series Segments Variance Accumulation for Time-series Segments 22 22

20 contd 23 23

21 contd 24 24

22 contd Scenario 1: RealTime Malware Detection (Dataset Composition 1x220 folds) Scenario 2: Standard - Cross Validation Scenario (Dataset Composition 10x22 folds) Classification Engine Information gain (refer TR) Machine Learning Classifier Decision Tree J48 Voting Method & Alarm 25 25

23 Results 26 26

24 Performance Overhead Processing Overhead (1685%) Feature Extraction + Classification Android - Performance Degradation (373%) 27 27

25 Conclusion A realtime analysis of time-series PCB dataset is presented to detect Malware A Novel scheme is used to extract hidden information in the execution patterns of the processes 110 Malware and 110 benign application are used to test the framework 99% DR with less than 1% Fp is achieved with realtime malware detection scenario A 90-93% Dr is achieved with 5-6% FP using standard 10 fold cross validation scenario Finally, 3-4% performance degradation of System is observed with framework All these benchmark make is this scheme suitable for realtime deployment 28 28

26 References Gartner, Gartner says asia/pacific led worldwide mobile phone sales to growth in first quarter of 2013, *last-viewed-june ], 2013 Trend-Micro, 3q 2012 security roundup: Android under siege: Popularity comes at a price, Research and Analysis, 2012 S HILL, Top 3 android security apps, do they protect you? on-december ), 2012 D Ehringer, The dalvik virtual machine architecture, Techn report (March 2010), 2010 I Witten and E Frank, Data mining: Practical machine learning tools and techniques, second edition Morgan Kaufmann, 2005 G Dini, F Martinelli, A Saracino, and D Sgandurra, Madam: a multi-level anomaly detector for android malware, in International Conference on Mathematical Methods, Models and Architectures for Computer Network Security, 2012, pp

27 References T Blasing, L Batyuk, A Schmidt, S Camtepe, and S Albayrak, An android application sandbox system for suspicious software detection, in International Conference on Malicious and Unwanted Software (MALWARE) IEEE, 2010, pp G Portokalidis, P Homburg, K Anagnostakis, and H Bos, Paranoid android: Versatile protection for smartphones, in Proceedings of the 26th Annual Computer Security Applications Conference ACM, 2010, pp I Burguera, U Zurutuza, and S Nadjm-Tehrani, Crowdroid: behavior-based malware detection system for android, in Proceedings of the 1st ACM workshop on Security and privacy in smartphones and mobile devices ACM, 2011, pp A Shabtai, U Kanonov, Y Elovici, C Glezer, and Y Weiss, Andromaly: a behavioral malware detection framework for android devices, Journal of Intelligent Information Systems, pp 1 30, 2011 Farrukh Shahzad, M Ali Akbar, Salman H, Khan, Muddassar Farooq, TStructDroid: Realtime Malware Detection using In-execution Dynamic Analysis of Kernel Process Control Blocks on Android, Technical Report [

28 Thank you Any questions? 31 31